HomeResearch
Model methodology · v12.4
DOC · RES-MOD-2026
IV.Research

The model. Audited.

Our projection model isn't a marketing trick. It's a weighted response estimate against an N=847 longitudinal cohort. Here's how it was built and how to verify it.

The cohort

847 subjects, ages 14–34, recruited between 2021 and 2025 across English-speaking online communities. Selection bias acknowledged: subjects were self-selecting and motivated. We do not generalize beyond that population.

  • Open-plate (14–18): n = 211
  • Fusing (18–22): n = 324
  • Closed-plate adult (22–34): n = 312
  • Mean adherence: 78% (median 84%)
  • Completion rate: 84% reached Week 12 rescan

The measurement

Baseline and Week-12 measurements were performed under standardized conditions: morning (within 15 min of waking), barefoot, against a fixed wall, with a 1mm-graduated stadiometer or equivalent. Self-measurement was validated against a clinician scan for ~10% of the cohort with <3mm variance.

The model

The projection model is a weighted multilinear estimate across four orthogonal axes — posture, endocrine baseline, genetic midparent gap, and lifestyle execution. Weights were fit against observed deltas in the n=520 calibration subset and validated on the held-out n=327 cohort.

  • against observed Δ when adherence ≥ 80%: 0.87
  • SEM on median projection: ±0.3″
  • Model card & weights: available on request to researchers
"This isn't a clinical trial. We don't claim it is. It's a longitudinal observational cohort study with a published response model. If you have a stronger methodology to offer, we will adopt it."— Methodology preface, v12.4

What we don't claim

  • Subjects past ~22 do not regrow long bones (open-plate growth).
  • The "+2.4″ median" includes posture & disc-height recovery, not just bone elongation. We are explicit about the mechanism in the FAQ and in every projection.
  • We have no randomized control. The placebo question is real and acknowledged.
  • Adherence below 80% produces unpredictable outcomes; we refund.

Open data & reproducibility

The anonymized dataset (n=847 with no PII) is available to researchers, journalists, and serious skeptics on signed request. We've also seeded the dataset with three independent reviewers (sleep researcher, biomechanist, statistician). Their notes are part of the model card.


How to get the dataset

Email [email protected] with affiliation and intent. We respond in 72 hours.

Audit it on yourself.

Run your 14-variable assessment. See the model output for your specific profile, side-by-side with the cohort.

Run my assessment